Out-of-Sample Testing
A strategy evaluation method that divides historical data into a development set and a validation set, where the validation set is used only once during final testing, in order to examine the generalization ability of the strategy.
Timeline (last 90 days)
Out-of-sample testing cuts the data in two, using one part for development and the other only once for final validation; walk-forward analysis proceeds on a rolling basis, training on window 1 and testing on window 2, then moving to training on window 2 and testing on window 3, and so on.
Walk-forward analysis is considered more resistant to overfitting than out-of-sample testing because it more closely resembles real trading, where old strategies are continually tested against new data.
All Facts (2)
Out-of-sample testing cuts the data in two, using one part for development and the other only once for final validation; walk-forward analysis proceeds on a rolling basis, training on window 1 and testing on window 2, then moving to training on window 2 and testing on window 3, and so on.
50%UnverifiedWalk-forward analysis is considered more resistant to overfitting than out-of-sample testing because it more closely resembles real trading, where old strategies are continually tested against new data.
50%